just checked, yes they don't seem to provide that information, most other providers are advertising fp8 or fp4 which is okay, but "together" doesn't, so they are likely using fp4
update: coreweave/fp8 is at 191 tps, launched this morning, but really bad cache hit rate (~60%), coreweave is good for privacy but let's hope they improve cache
Of course. This is just a list of someone's ideas for equivalently difficult (and transformative) unsolved problems in biology. Many of these would of course have obvious and immediately impactful practical applications, unlike the math problems.
> ... This is just a list of someone's ideas for equivalently difficult (and transformative) unsolved problems in biology. ...
I think GP was saying the proof of the pudding is in the tasting: the longer a problem has provably resisted resolution the more difficult it is considered...
bombastically decorating a problem as equivalently difficult does not make it so.
Well, if LLMs actually turn out to solve some of the Millennium problems, at least some of these biology problems are almost certainly at least as difficult. Certainly solving any of them would be Nobel-worthy if biology nobels existed (and solving some of them would likely easily be worthy of a medicine Nobel). It's not like these are new problems whose difficulty is unknown!
The Millenium Prize problems didn't only withstand decades/century of resolution, they also withstood the same test of time in ridicule of the challenge statement.
This list of "Millenium Problems for Biology" contains such brainfart level "analogies" that there the list will be ridiculed, for the question / challenge itself displays a lack of understanding of the subject in question. Science is also asking the right questions.
Consider for example:
> 10. Protein Amplification Chain Reaction:
> Demonstrate exponential amplification of arbitrary peptide substrates.
> Specifically, demonstrate input-protein-dependent synthesis of new, full-length, sequence-faithful covalent polypeptide copies from amino-acid monomers without a nucleic-acid template or preformed cognate scaffold, in a single pot reaction. For the challenge to be considered complete, at least 100 random peptide sequences of at least 50 amino acids each must be preregistered, synthesized, and pooled. It must then be shown that the abundance of these peptides in solution can be amplified at least 1000x with at least 90% sequence accuracy on a per-residue basis. Reasonable modifications may be added to the peptide sequences to facilitate post-amplification analysis if necessary, provided they are not active in the amplification. Methods that rely on explicit sequencing of the peptide are not permitted. Methods that rely on reverse translation to generate a nucleic acid intermediate are not permitted, because they are duplicative with a separate Millennium Problem.
The analogy is very clear: to amplify DNA or RNA one uses PCR, basically throw the desired product in a cauldron with monomer building blocks, then by repeated heating and cooling the lone monomers find their permitted locations on a complementary pre-existing strand, and form the new polymer strand.
So it seems natural to ask for a generalization to protein polymers, except every biologist or chemist knows its nonsense: proteins don't have a complementary strand! You can't demand chemistry or physics to magically copy without a complementary template!
You may ask "but if that were true, how can we already have PCR for RNA?"
Well pretty simple: while this is done routinely, its only possible indirectly: convert the RNA to double-strand DNA, use PCR on this DNA and then convert the amplified DNA back to RNA!
The demand to not involve sequencing or the hypothetical reverse translatase from one of the other problem statements turns this one into a non-existence theorem, but the challenge doesn't describe a winner for demonstrating its impossibility!
I assure you that any chemist or biologist being asked why we dont have PCR for protein, will understand your lack of knowledge, and explain how PCR works, so that you understand that PCR was only possible because of the complementary strand!
This list will be ridiculed for being not even wrong.
You can then run your script/dev with: sops exec-env secrets.enc.env 'docker xxxx' (it will ask you for your password, or touch-id to decrypt the secrets)
I like this because this way the .env doesn't sit in the directory at all, and is only passed to your dev environment and stays in it while it's running
Decrypt back to a file (if you ever want that): sops decrypt secrets.enc.env > .env
---
Well ofc, any agent can do docker inspect to get all those env vars, but atleast reading the dotfiles won't do anything
you can also edit the file with: sops --input-type dotenv --output-type dotenv secrets.enc.env
This is crazy, but sadly anthropic/openai will never do this, what has happened to this world, where chinese companies are more open than US or even EU companies
I don’t think Trump changed them, but Trump is absolutely a symptom of larger social collapse in the US, and that collapse has affected Altman and Amodei. We’re not even pretending that truth matters or that the wealthy can ever suffer consequences, and those two seem quite liberated by that.
"As we get closer to building AI, it will make sense to start being less open. The Open in OpenAI means that everyone should benefit from the fruits of AI after its built, but it's totally OK to not share the science (even though sharing everything is definitely the right strategy in the short and possibly medium term for recruitment purposes)."
-Ilya Sutskever (email to Elon musk and Sam Altman, 2016)
I think this is more about the UI, rather than the install. I haven't tried it myself though.
I think nowadays using quickshell anyone who is so inclined can vibecode their own UI though. I recently made the switch to Wayland/hyprland and rebuilt my polybar on quickshell, even adding widgets that allow getting system info/fine grained system control (interactive Bluetooth, WiFi, Volume, Brightness etc).
Even if you introduce bugs in your UI (which I think is not very likely if you have a basic understanding of your system), the chance that someone would exploit software that literally only runs on your own machine seems extremely unlikely to me.
Yeah, I could literally open up a port on my computer which executes everything sent to it. As a bespoke app, it's pretty unlikely that it would ever be exploited even though it's ridiculously dangerous.
To be at risk, you'd first need to publish your application. The attacker would separately have to figure out how to identify and access your computer for attack.
The old human nature tendency of "Just ship it!" and the desire to move on to the next shiny thing on our wishlist doesn't go away with AI. If it anything, it makes it worse.
The problem is not the hype, the problem is that it's a vibe coded distro. AI lowers the bar substantially in terms of needed competence and diligence to create something like this. Of course bugs and issues are present in human created software as well, but now that everyone can vibe code a distro the chances of low quality software are much greater.
Add that annoying theo guy to that list. Cant stand these people, they confidently push out videos like they're experts, a week later it turns out whatever they were talking about was total crap and they've already abandoned it - case in point OpenClaw. Look at the mess of videos those named above put out about it, not a single one uses it anymore.
Sadly, true. I know a person who gets AI news from that YouTuber and quotes him like he’s a household name and pays him for some repackaged chat models. It’s tedious and hard to talk to them about basic/remedial aspects because their education from YouTube is half-heard and superficial. I’m trying to get him to invest the time in hands-on experience and then we can talk about that. When you’re deep into these channels, you’re not gaining experience and you can’t easily start until you pull away.
There is only so much a human can master in his lifetime. And if you choose to master the art of video production, then you are probably not spending that much time on mastering the thing you yap about on camera…
> if you choose to master the art of video production, then you are probably not spending that much time on mastering the thing
What?
Its perfectly possible to master two things. Video production isnt _that_ hard. Especially as lighting, shooting adding graphics and editing film is much much easier than 10 years ago.
I was watching one guy explaining the xv6 code and it was pretty much one shot with a single camera pointing at the listing on paper. Not sure if you need professional video if you want to really expound on something.
I don't think I've ever actually seen a theo video, but I stumbled across this and think I've seen all I need to know. 3:45 even references your criticism.
It’s sad, but a lot of software folks have decided to become influencers and not pursue expertise. I remember before you’d mostly be reading boring looking blogs to follow experts in the industry. It seems like the mindshare has moved to flashy videos by people that aren’t experts, but are great at communication. Similar to how if you want to watch some diy video on YouTube, you’ll get much better information from a poorly produced snd edited video from someone that knows what they’re talking about vs a well produced and edited video from someone that has no idea what they’re talking about.
> they confidently push out videos like they're experts, a week later it turns out whatever they were talking about was total crap
100%! When the US attacked Iran, they became experts on wars; when Iran closed the strait, they turned experts on foreign policy, diplomacy, and negotiation; they were doctors during COVID; now they are SME in flooding and disaster control after the Nepal incident.
UI is desktop environment and (usually) ports to large-use distros cleanly.
- JaKooLit’s Fedora-Hyprland Repository: https://github.com/JaKooLit/Fedora-Hyprland. The most popular automated setup guide and installer for Fedora, bundling Hyprland alongside pre-configured bars, launchers, and
- Official Hyprland Wiki: wiki.hypr.land/Getting-Started/Installation/. The main reference guide for core configuration options, environment variables, and Wayland portal requirements.
- Solopasha Fedora COPR copr.fedorainfracloud.org/coprs/solopasha/hyprland. The primary community repository hosting cutting-edge builds of Hyprland and its ecosystem packages for
- Fedora Discussion Tutorials: discussion.fedoraproject.org. Community walkthroughs covering minimal netinstall setups and distro-specific Wayland troubleshooting.
Fedora (or debian, or archlinux or whatever distro you choose) doesn't force you into an UI. All of them have minimal, server and/or "flavored" versions and you can always uninstall and install whatever you want.
I run Niri and Dank Material Sell on Fedora. It’s perfect, and better than Omarchy from what I’ve seen. I much prefer scrolling window managers to tilers.
Yup. I've been on zirconium (https://github.com/zirconium-dev/zirconium) for a while, and it's great. Fedora atomic/bootc under the hood with Niri+DMS. Basically bombproof.
You still have to use rpmfusion, which isn’t a huge deal. It’s maybe 5 min one time and then you are set.
Or you could use one of the Universal Blue spins. Bluefin and Aurora have non-free stuff built in and you can install pretty much any package from anywhere.
Counterpoint: I've had very few (and always minor) problems upgrading between major Fedora releases over the years with no more effort than following the well-documented steps to do so, including both traditional and atomic variants, the latter with a wide variety of layered packages as specifically not recommended due to potential problems when upgrading.
And on the server side of things, I ran a Fedora CoreOS VM with a Pihole container that silently and successfully auto-updated both the OS and the service container for years with no intervention on my part.
I had to support fedora on desktop and laptop for a large company.
Worse, it wasn't my job to support those machines, its just my department needed cuda and I was the slowest person to run away when they called for help.
I had just enough issues with archinstall the last time I tried it that I went back to EndeavourOS which essentially gets you a GUI installer and some pre-installed utilities on the stock Arch install. The only significant change I'm aware of is that they use dracut instead of mkinitcpio.
I think I get your point as omarchy represents a different kind of UI, but if it was just about getting a rock solid stable working environment, Fedora or Ubuntu with either Gnome and KDE work perfectly fine. I don't even use a third party terminal (just konsole). Keyboard driven terminal workflows works fine on KDE for me.
> As a developer who mostly wants a Mac-like Linux distro without the associated noise
I'm required to use Ubuntu at work. Coming from Mac, apart from the menu bar at the top, on Gnome, I've been able to customize the keyboard shortcuts, remap the keyboard so that Ctrl works like Cmd, and use extensions like Dash to Dock[1] and themes like Whitesur[2] to replicate something that almost works like a Mac.
The keyboard remapping and customizing keyboard shortcuts were all done within default the default settings app.
The only things missing are some keyboard shortcuts like Ctrl+A/V to move to the beginning and end, and the Ctrl+Shift+C/V behavior on the terminal instead of Cmd+C, which I've just worked around by using VSCode's terminal and configuring it to copy when I press Ctrl+C with some text selected.
> The only things missing are some keyboard shortcuts like Ctrl+A/V to move to the beginning and end, and the Ctrl+Shift+C/V behavior on the terminal instead of Cmd+C, which I've just worked around by using VSCode's terminal and configuring it to copy when I press Ctrl+C with some text selected.
The fact that you think this is acceptable shows that you don't appreciate the difference. Again, I haven't used Omarchy but the docs claim that Super + C and Super + V work everywhere, not with all the different exceptions you mention.
if you reset the keyboard shortcuts to their defaults and install toshy you can get the ctrl-a/e stuff and copy and paste in the command line with cmd-c/v and a lot of the other mac like keyboard shortcuts. it's really plug and play and works well... can't recommend it enough. (though there are some edge cases that may require some tweaks but overall it's great)
Homebrew is like giving a bunch of wikipedia randos remote shell access to your computer. There is no enforced code review policy. Any maintainer can make a commit under a pseudonym and merge their own code to main without review.
Homebrew should not be allowed near any machine you need to be able to trust.
I love wikipedia for research. It is great that anon randos can help keep a common encyclopedia maintained with high accountability. If a mistake happens it is quickly corrected.
But that trust model does not work with software.
It is negligent that MacOS users, even those with prod access, all let any anon randos that sign up to be Homebrew maintainers execute any code they want on their computers with no accountability. By the time someone figures that out and corrects it, the damage is done.
What is worse, is almost zero MacOS users I talk to are aware of this risk. Even security engineers. Many say "no worse than NPM" as if that is a defense.
Only honest commits by good actors are reviewed. Good for QA but useless for security.
Cryptographic signing is not enforced for commits or reviews, so an attacker that controls a single maintainer Github API key could make a PR with a burner account then "review" and merge their own PR. Time it right and bury it in a dependency of a dependency and you are likely to get away with it. Especially considering they also do not do full source bootstrapping or enforce deterministic builds so including random binary-blobs in tree is a-okay and not sus at all.
The fact everyone is not aware of this is exactly why it is going to work, or already has.
I did once propose to the Homebrew community to enforce commit and review signing but they ultimately refused, worrying that asking people to press a button when it blinks would cause too many definitely smart and useful contributors to leave the project.
> In a different way you can see this in this thread where instead of discussing why Omarchy has the bug
There's nothing interesting to discuss. Rootless OCI-compliant containers exist, and Omarchy ignored them. The "convenience" of the solution they chose simultaneously opens an enormous attack surface that the maintainers didn't consider.
This is why I don't believe the "give me a macOS distro" people - even Apple wouldn't do this. If you want a preconfigured Mac-like distro, then you should use GNOME and not a pre-riced desktop with hundreds of dotfiles. You don't have to use Arch either, you can go with a graphical install of Fedora or CachyOS if you want. This will give you a system that you can understand, maintained by people with a minimum standard of quality, that actually resembles the workflow of macOS. There is no Niri or Sway rice that will magically make your system make sense, and this is why I think a lot of the Mac and Windows expats should just use a normal desktop.
> There's nothing interesting to discuss. Rootless OCI-compliant containers exist, and Omarchy ignored them.
Why Omarchy chose to ignore all the rootless OCI-compliant containers; and also why docker itself isn't rootless yet is a pretty interesting thing to discuss. I'm not aware of the specifics being a Mac user myself, but others in the thread (e.g. here https://news.ycombinator.com/item?id=49500186) say that podman still doesn't work 1:1 with docker exactly.
If you disagree, that's a reasonable thing to discuss; but otherwise just saying "ha ha just use the rootless thing" is missing the point.
If you're not going to use rootless containers, then you simply cannot ship the rootful solution as default in a desktop operating system.
There's absolutely meat on the bones regarding root/rootlessness, but that's not the problem at-hand. The problem is shipping vulnerable defaults, and the case is closed.
> if you encounter a problem with Arch and ask about it
It's probably already documented on Arch wiki or it's a generic Linux issue that has already been solved by someone (that has been my experience with Cachy).
No harm, no foul izacus. I am laughing inside because I'm included in the "angry" group by default (I hate all Linux distros, less than Windows or MacOS) and I'm often a basic docker user but got over major DE customization like Omarchy when I needed to get thing done at a higher rate (back then the coolness was crunchbang).
Omarchy imo is best for agent maxxing Mac power users who aren't locked into Apples proprietary apps like Final Cut, audio production software or Adobe and Affinity, so maybe not so great for designers and photographers as the main system.
It's also half-coded in Javscript, with a grand total of 881 unique dependencies (271 npm + 614 Rust; derived from 38 directly declared dependencies). What could possibly go wrong?!
I never thought about wanting an HN comment as a wearable shirt before, but this ones one I would buy. But seriously, I love Claude and building all sorts of projects, but something as crucial as a password manager is a little bit too risky.
Reminds me of letting an LLM generate a strong password in the first place. Very practical, very quick, no terminal commands needed! I'm sure some have done exactly this. But then you realise your password is always shared with your LLM provider, and it's not so random after all (unless it used tools).
I don't have an issue with a vibe coded password manager, LLMs are insanely good. The issue, atleast to me, is the UI is extremely low effort LLM slop. I'd have appreciated something native that was super fast.
reply